Add details of multi-screen origin support to relevant features #21123

chrisdavidmills · 2023-10-30T13:35:24Z

Summary

The Window Management API was implemented in Chrome 100 (see ChromeStatus entry). I originally submitted BCD updates for this in #20600.

However, during the tech review of my docs additions for this API, I added some details of how the multi-screen origin concept introduced with this API affects the behavior of some existing related features. This was done on the instruction of the tech reviewer, @michaelwasserman.

In his editorial review, @wbamberg then rightly pointed out that this information belongs in the BCD. This PR adds the relevant information to BCD. The general pattern is:

Chromium supports the relevant features relative to the multi-screen origin, as it supports the Window Management API, but it also seemed to support a non-standard form of the multi-screen origin, at least for some features, before that.
Firefox supports its own non-standard version of the multi-screen origin, for use with these features, but its implementation is buggy/inconsistent
In most cases, Safari supports the relevant features relative to the current or primary screen. However, it does report the Window properties (e.g. screenX) relative to a multi-screen origin.

Does this sound accurate, @michaelwasserman?

It would be nice to have more accurate, detailed information on the exact behavior in Firefox, but I'm not sure how to get this without spending a lot of time doing detailed testing. At least this way, readers are warned about the buggy behavior.

Test results and supporting details

Related issues

michaelwasserman

It'd take more testing (and better WPTs!) to be really thorough here, but in general: Firefox seems to support relative-multi-screen coordinates fairly well, while Safari has some inconsistencies around window.open(). Hopefully my comments are helpful.

michaelwasserman · 2023-10-30T22:52:33Z

api/Screen.json

+              "chrome": {
+                "version_added": "1"
+              },
+              "chrome_android": "mirror",


I'm uncertain about any Chrome behavior on Android's multi-screen freeform windowing mode. Ditto for anything else mobile... sorry.

A part of me thinks that this is perhaps not so important, as surely not many people are going to be using mobiles with multiple screens...but I'm really not sure. When I tried my demo app on Chome Android, it didn't seem to work (the windows wouldn't open), although the feature detection returned true (i.e. it is supported).

I wonder if the BCD collector app can give us anything interesting with regards to mobile support?

michaelwasserman · 2023-10-30T23:11:57Z

api/Screen.json

+              "firefox": {
+                "version_added": "1",
+                "partial_implementation": true,
+                "notes": "Behavior is inconsistent/unreliable."


I'd say Firefox has full implementation support here, and for most API surfaces representing coordinates relative to the multi-screen origin. In general, I have observed inconsistent/unreliable behavior at times in the past, but never fully cataloged them, and Firefox 115 ESR on Linux X11 and Firefox 118 on Mac seems coherent in a few minutes of basic testing: screen.left|top, window.moveTo(), window.open(), and window.screenX|Y use multi-screen origin coordinates. moveTo|By clamps windows to their current display, but window.open can open a popup with coordinates on any display.

OK, thanks. I've updated the data points to represent this information.

michaelwasserman · 2023-10-30T23:46:13Z

api/Window.json

+              "firefox": {
+                "version_added": "1",
+                "partial_implementation": true,
+                "notes": "Behavior is inconsistent/unreliable."


Ditto, I'd say Firefox has full implementation support for specifying moveTo and open coordinates relative to the multi-screen origin, even if requests are clamped to the current screen sometimes (but not others).

OK, I've put that for moveTo/By, multi-screen coordinates are used, but windows are clamped to their existing displays. For open(), I've just put no note, i.e. it works perfectly for multi-screen origin. That seemed to be what you were saying above?

michaelwasserman · 2023-10-30T23:58:35Z

api/Window.json

+              "opera_android": "mirror",
+              "safari": {
+                "version_added": false,
+                "notes": "Opened relative to the origin of the primary screen."


I can't figure out Safari's behavior here... in one test, top was relative to the current screen, but left from an opener on a secondary screen was completely disregarded, then I rearranged displays and left was relative to the primary screen, while top from an opener on a secondary screen was completely disregarded. That's getting into territory where I'd use terms like inconsistent and unreliable....

Crikey, the Safari comments are not very encouraging.

I'm not certain it is really worth trying to understand and document the exact behavior of Safari, as it is so buggy and weird. For the moment I've just put the "Behavior is inconsistent/unreliable" note on the Window features for Safari, and left it as support false on the Screen features (because the original data said false for Safari, before I started working on this API).

Thanks again for all your time and effort spent helping me to understand this stuff.

michaelwasserman · 2023-10-31T00:07:05Z

api/Window.json

+              "opera_android": "mirror",
+              "safari": {
+                "version_added": false,
+                "notes": "Opened relative to the origin of the primary screen."


Safari 17's behavior here is closer to 'inconsistent' or 'unreliable'; it seems to depend on the relative arrangement of displays. For example:

Screens arranged primary to the left/right of secondary - left|top are relative to primary

Screens arranged primary above/below secondary - left|top are relative to the opener's current screen

It gets even weirder if the displays don't share a top or left edge, e.g.:

Screens arranged primary diagonal below+left of secondary - left is relative to primary, but top is (erroneously clamped?) to the top of the primary screen.

Screens arranged primary diagonal above+left of secondary - left is relative to primary, but top is (erroneously clamped?) to the bottom of the primary screen.

michaelwasserman · 2023-10-31T00:20:34Z

api/Window.json

+              "oculus": "mirror",
+              "opera": "mirror",
+              "opera_android": "mirror",
+              "safari": {


Seems like Safari's screen.top and window.screenY is always relative to itself, but screen.left and window.screenX are relative to the multi-screen origin... it's hard to understand...

michaelwasserman

lgtm with some minor comments, thanks for sorting through the unfortunate complexity here.

michaelwasserman · 2023-10-31T17:05:33Z

api/Screen.json

+              "standard_track": false,
+              "deprecated": false
+            }
+          }
        }
      },
      "availWidth": {


Should edge be mirror here? (line 224)

The trouble here is that if we mirror it, it will show the equivalent to Chrome 1, which is wrong, as Edge wasn't based on Chromium at that time. I also tried adding in an extra piece of support data to show that later on, Edge 100 (equivalent to Chrome 100) supports this without the note, but then it wants me to call the earlier support a partial implementation. But that's not really right either, as the old Edge implementation was correct that the time.

So I've opted to just leave it alone.

michaelwasserman · 2023-10-31T17:11:49Z

api/Window.json

@@ -2789,6 +2789,42 @@
            "standard_track": true,
            "deprecated": false
          }
+        },
+        "relative-multi-screen": {


I think we can drop this section, moveBy's parameters are relative offsets from the window's current position, not absolute coordinates in a single- or multi-screen space.

OK, makes sense. Removed.

michaelwasserman · 2023-10-31T17:13:20Z

api/Window.json

+              "opera": "mirror",
+              "opera_android": "mirror",
+              "safari": {
+                "version_added": "1",


Is it possible to leave this marked false with the note? Otherwise, this is probably okay.
Ditto for all other Safari entries.

Your suggestion doesn't sound correct to me — if it isn't supported at all, how can it exhibit inconsistent/unreliable behavior?

Understood, this framing is reasonable.

michaelwasserman · 2023-10-31T17:16:58Z

api/Window.json

+              "edge": "mirror",
+              "firefox": {
+                "version_added": "1",
+                "notes": "Multi-screen coordinates are used, but windows are clamped to their existing displays."


I might either omit this, or note similar clamping in Chrome without window-management permission, and ditto for Chrome's window.open().

Notes added as requested:

Without the window-management permission, multi-screen coordinates are used, but windows are clamped to their existing displays.

michaelwasserman · 2023-10-31T17:21:32Z

api/Window.json

-              "version_added": "1"
+              "version_added": "1",
+              "partial_implementation": true,
+              "notes": "Behavior is inconsistent/unreliable."


Should screenX have a similar note?

Ooop, the issue here is that I've got the note on the screenY main feature as well as the relative-multi-screen subfeature, whereas it should just be on the subfeature. I've fixed that now.

michaelwasserman

lgtm, ty!

Elchi3

One suggestion otherwise I think this is good!

Elchi3 · 2023-11-03T12:07:36Z

api/Screen.json

+              "safari": {
+                "version_added": "1",
+                "partial_implementation": true,
+                "notes": "Behavior is inconsistent/unreliable."


Is "behavior" a good word here? Sounds very generic to me. Could we instead say "Origin relation" to be a bit more precise what exactly is off here? (if you agree, please update throughout).

Suggested change

"notes": "Behavior is inconsistent/unreliable."

"notes": "Origin relation is inconsistent/unreliable."

@Elchi3 I think you've got a point, but it feels to me like "Origin relation" is a bit difficult to comprehend. How about "Coordinate origin used is inconsistent/unreliable"?

Sure, that works, too! :)

Cool, notes updated!

…21123) * Add details of multi-screen origin support to relevant features * Making some changes based on michaelwasserman comments * Fixes for 2nd round of michaelwasserman comments * Update Safari notes as discussed

Add details of multi-screen origin support to relevant features

0787b34

github-actions bot added the data:api Compat data for Web APIs. https://developer.mozilla.org/docs/Web/API label Oct 30, 2023

chrisdavidmills mentioned this pull request Oct 30, 2023

Editorial review for Window Management API docs mdn/content#29719

Merged

michaelwasserman reviewed Oct 31, 2023

View reviewed changes

chrisdavidmills added 2 commits October 31, 2023 15:10

Merge branch 'main' into multi-screen-origin-updates

29153e7

Making some changes based on michaelwasserman comments

9b16687

michaelwasserman approved these changes Oct 31, 2023

View reviewed changes

Fixes for 2nd round of michaelwasserman comments

dae5def

michaelwasserman approved these changes Nov 1, 2023

View reviewed changes

Elchi3 reviewed Nov 3, 2023

View reviewed changes

Update Safari notes as discussed

4879d37

chrisdavidmills requested a review from Elchi3 November 6, 2023 08:28

Elchi3 approved these changes Nov 6, 2023

View reviewed changes

Elchi3 merged commit 07a561b into mdn:main Nov 6, 2023
4 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add details of multi-screen origin support to relevant features #21123

Add details of multi-screen origin support to relevant features #21123

chrisdavidmills commented Oct 30, 2023

michaelwasserman left a comment

michaelwasserman Oct 30, 2023

chrisdavidmills Oct 31, 2023

michaelwasserman Oct 30, 2023

chrisdavidmills Oct 31, 2023

michaelwasserman Oct 30, 2023

chrisdavidmills Oct 31, 2023

michaelwasserman Oct 30, 2023

chrisdavidmills Oct 31, 2023

chrisdavidmills Oct 31, 2023

michaelwasserman Oct 31, 2023

michaelwasserman Oct 31, 2023

michaelwasserman left a comment

michaelwasserman Oct 31, 2023

chrisdavidmills Nov 1, 2023 •

edited

Loading

michaelwasserman Oct 31, 2023

chrisdavidmills Nov 1, 2023

michaelwasserman Oct 31, 2023

chrisdavidmills Nov 1, 2023 •

edited

Loading

michaelwasserman Nov 1, 2023

michaelwasserman Oct 31, 2023

chrisdavidmills Nov 1, 2023

michaelwasserman Oct 31, 2023

chrisdavidmills Nov 1, 2023

michaelwasserman left a comment

Elchi3 left a comment

Elchi3 Nov 3, 2023

chrisdavidmills Nov 3, 2023

Elchi3 Nov 3, 2023

chrisdavidmills Nov 6, 2023

	"notes": "Behavior is inconsistent/unreliable."
	"notes": "Origin relation is inconsistent/unreliable."

Add details of multi-screen origin support to relevant features #21123

Add details of multi-screen origin support to relevant features #21123

Conversation

chrisdavidmills commented Oct 30, 2023

Summary

Test results and supporting details

Related issues

michaelwasserman left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

michaelwasserman left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

chrisdavidmills Nov 1, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

chrisdavidmills Nov 1, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

michaelwasserman left a comment

Choose a reason for hiding this comment

Elchi3 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

chrisdavidmills Nov 1, 2023 •

edited

Loading

chrisdavidmills Nov 1, 2023 •

edited

Loading